Psychological Science
http://pss.sagepub.com/
On the Nature and Nurture of Intelligence and Specific Cognitive Abilities: The More Heritable, the More
Culture Dependent
Kees-Jan Kan, Jelte M. Wicherts, Conor V. Dolan and Han L. J. van der Maas
Psychological Science 2013 24: 2420 originally published online 8 October 2013
DOI: 10.1177/0956797613493292
The online version of this article can be found at:
http://pss.sagepub.com/content/24/12/2420
Published by:
http://www.sagepublications.com
On behalf of:
Association for Psychological Science
Additional services and information for Psychological Science can be found at:
Email Alerts: http://pss.sagepub.com/cgi/alerts
Subscriptions: http://pss.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
>> Version of Record - Dec 13, 2013
OnlineFirst Version of Record - Oct 8, 2013
What is This?
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
493292
research-article2013
PSSXXX10.1177/0956797613493292Kan et al.Nature and Nurture of Intelligence
Research Article
On the Nature and Nurture of Intelligence
and Specific Cognitive Abilities: The More
Heritable, the More Culture Dependent
Psychological Science
24(12) 2420–2428
© The Author(s) 2013
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797613493292
pss.sagepub.com
Kees-Jan Kan1,3, Jelte M. Wicherts2,3, Conor V. Dolan1,3, and
Han L. J. van der Maas3
1
Department of Biological Psychology, VU University; 2Department of Methodology and Statistics,
Tilburg University; and 3Department of Psychological Methods, University of Amsterdam
Abstract
To further knowledge concerning the nature and nurture of intelligence, we scrutinized how heritability coefficients
vary across specific cognitive abilities both theoretically and empirically. Data from 23 twin studies (combined N =
7,852) showed that (a) in adult samples, culture-loaded subtests tend to demonstrate greater heritability coefficients
than do culture-reduced subtests; and (b) in samples of both adults and children, a subtest’s proportion of variance
shared with general intelligence is a function of its cultural load. These findings require an explanation because they
do not follow from mainstream theories of intelligence. The findings are consistent with our hypothesis that heritability
coefficients differ across cognitive abilities as a result of differences in the contribution of genotype-environment
covariance. The counterintuitive finding that the most heritable abilities are the most culture-dependent abilities sheds
a new light on the long-standing nature-nurture debate of intelligence.
Keywords
intelligence, behavior genetics, cognitive ability, environmental effects, individual differences
Received 5/7/12; Revision accepted 5/17/13
Whether intelligence depends more on nature or on nurture is a long-standing issue dating back to 17th-century
rationalism and empiricism and with roots in the ancient
Greek philosophies of Plato and Aristotle (Fancher, 1996).
With the emergence of psychometrics and behavioral
genetics in the 20th century, it became possible to address
this issue empirically—in terms of individual differences—through the decomposition of variance in psychometric intelligence into genetic and environmental
variance components (Plomin, DeFries, McClearn, &
McGuffin, 2008). At first sight, the results seem to favor
nature: In samples of adults, on average, the genetic variance components account for up to 80% of the variance
in full-scale IQ and general intelligence (Plomin et al.,
2008). However, the results on which this average is
based were often derived under assumptions that may
not have been met. For example, in behavior-genetic
models of intelligence, genotype and environment are
commonly assumed to be independent and, hence, do
not covary, whereas genotype-environment covariance is
presumably present and might account for as much as
30% of the variance in adult IQ ( Johnson, Penke, &
Spinath, 2011).
Our aim in the present research was to further knowledge concerning the nature and nurture of intelligence
by scrutinizing how heritability coefficients vary across
specific cognitive abilities, both theoretically and empirically. We evaluated the implications of the empirical findings for theories of intelligence, notably, the mainstream
g theory ( Jensen, 1998; Spearman, 1927) and fluid-crystallized theory (Cattell, 1971).
Corresponding Author:
Kees-Jan Kan, Department of Biological Psychology, VU University,
Van der Boechorststraat 1, Amsterdam 1081 BT, The Netherlands
E-mail: [email protected]
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
Nature and Nurture of Intelligence
2421
How Do Heritability Coefficients Differ
Across Cognitive Abilities?
Theory
In both g theory and fluid-crystallized theory, intelligence
is conceptualized as a major, largely genetically fixed
source of individual differences in IQ (referred to as g,
for general intelligence, in g theory; as Gf, for fluid intelligence, in fluid-crystallized theory; and henceforth in
this article, as g). On the basis of these theories, cognitive
abilities and IQ subtests are often categorized as fluid or
crystallized. IQ subtests are further viewed as being—to
varying degrees—culture reduced or culture loaded
( Jensen, 2012).
Fluid abilities are assessed by subtests that minimize
the role of individual differences in prior knowledge
(henceforth, fluid tests), and crystallized abilities by subtests that maximize it (henceforth, crystallized tests).
Individual differences in fluid-test scores thus reflect primarily the sources of individual differences in on-thespot cognitive processing (e.g., reasoning or memorizing),
whereas individual differences in crystallized-test scores
reflect primarily the sources of individual differences in
previously acquired knowledge and skills. Because
knowledge and individual and group differences in
knowledge are strongly culturally influenced, crystallized
tests require relatively many adjustments to adapt them
from one language or culture to the next (Georgas, van
de Vijver, Weiss, & Saklofske, 2003). In this sense, crystallized tests are typically more culture loaded than are fluid
tests.
From both g theory and fluid-crystallized theory, it follows that heritability coefficients differ across specific
cognitive abilities and IQ subtests. In g theory, this can be
deduced from the following assumptions ( Jensen, 1987,
1998). First, individual differences in IQ scores reflect the
sources of individual differences in cognitive processing,
either directly (in fluid tests) or indirectly (in crystallized
tests). Second, among these sources, g is the most heritable source. Third, individual differences in IQ-subtest
scores reflect individual differences in g to different
degrees: The more complex an IQ test is (i.e., the more
cognitively demanding solving its items is), the more
individual differences in subtest scores reflect the relative
contribution of g. From this it follows that, compared
with noncomplex fluid tests (e.g., forward digit-span
tests), complex fluid tests (e.g., abstract-reasoning tests)
have relatively strong relations to g, as indicated by their
g loadings, and thus display relatively high heritability
coefficients.
The level of cognitive demand required to solve the
items of crystallized tests is considered to be relatively
low ( Jensen, 2012). Hence, to account for the substantial
g loadings of crystallized tests, theorists must invoke one
or more additional assumptions. In this regard, it is common for g theorists ( Jensen, 1998) to adapt the investment hypothesis, as formulated in fluid-crystallized theory
(Cattell, 1971), from which it also follows that heritability
coefficients differ across abilities and subtests.
The investment hypothesis holds that the acquisition
of knowledge depends strongly on cognitive processing,
such that individual differences in acquired knowledge
largely reflect fluid abilities and, thus, the underlying
sources of individual differences therein, notably, g.
Ultimately, individual differences in crystallized abilities
and crystallized-test scores largely reflect the same
genetic and environmental variables as individual differences in fluid abilities. However, given that additional
(non-g) influences (e.g., education) play a role during the
acquisition of knowledge, the investment hypothesis
holds that in the general population, the heritability coefficients of crystallized abilities will be lower than those of
fluid abilities (Cattell, 1971). Yet if the variance in the
additional influences is small—for example, in culturally
homogenous samples—the heritability coefficients of
crystallized abilities are expected to approach the heritability coefficients of fluid abilities ( Jensen, 1998).
In the absence of other hypotheses and, hence, on the
basis of the subtest-complexity and investment hypotheses alone, heritability coefficients of crystallized abilities
are expected not to exceed those of fluid abilities (see
the left panel of Fig. 1). Whether empirical findings (e.g.,
Pedersen, Plomin, Nesselroade, & McClearn, 1992) support this expectation has been questioned (e.g., Horn,
1985; Mackintosh, 1998), but to date, definite results
are scarce. Moreover, the consideration of the role
of genotype-environment covariance (e.g., Scarr &
McCartney, 1983) in the development of intelligence may
call for a revision of this expectation, as we outline next.
Heritabilities of IQ and g increase gradually over the
course of the life span (Haworth et al., 2010). This phenomenon has been attributed to a gradual increase in
active genotype-environment covariance (Haworth et al.,
2010; Johnson et al., 2011), which is thought to arise
because people with relatively high levels of cognitive
ability increasingly actively seek out and, therefore, are
exposed to cognitively stimulating environments (Dickens
& Flynn, 2001; Haworth et al., 2010; Johnson et al., 2011;
Scarr & McCartney, 1983).
The relative contribution of genotype-environment
covariance may differ across abilities. If stimulating environments foster societally valued knowledge and skills
more than cognitive processing per se, we expect, on the
basis of computer simulations with dynamical models
(Dickens, 2008; van der Maas et al., 2006), that (a) individual differences in culture-loaded tests should be relatively strongly related to g and (b) heritability estimates of
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
Kan et al.
2422
Subtest-Complexity and Investment Hypotheses
Dynamical Gene-Environment-Interplay Hypothesis
Sp
n
t
e
eD
iv
nit
ma
he
Crystallized Tests
nt
i
nd
s
Pa
g
Co
Proportion of Variance Explained by the
Most Heritable Major Source (g or Gf)
CultureReduced
Tests
Heritability Coefficient (h 2)
Noncomplex
Fluid Tests
ot
iv
nit
g
Co
CultureDependent
Tests
do
ma
e
eD
he
nt
Crystallized (Knowledge) Tests
Cultural Load
Heritability Coefficient (h 2)
Complex Fluid Tests
nt
e
nm
iro
nv
ty
no
cts
e
Eff
-E
pe
e
fG
o
ist
tat
S
ha
l
ica
ce
an
ri
va
Co
tor
ac
F
ral
ne
Ge
it
nW
io
lat
Re
Fluid (Cognitive-Processing) Tests
Societal Demand
Proportion of Variance Explained by Less
Heritable (Non-g or Non-Gf) Sources
Fig. 1. Schematic illustration of the subtest-complexity and investment hypotheses from mainstream theories of intelligence, which predict that
heritability coefficients of culture-loaded crystallized abilities do not exceed those of culture-reduced fluid abilities (left panel), and our dynamical
gene-environment-interplay hypothesis, which predicts relatively high heritability coefficients of culture-loaded crystallized tests (right panel). g =
general intelligence; Gf = fluid intelligence.
culture-loaded knowledge tests should be affected relatively strongly by genotype-environment covariance,
which should ultimately result in the heritability coefficients of culture-loaded knowledge tests exceeding those
of tests that measure cognitive-processing abilities (see
right panel of Fig. 1).
The way in which heritability coefficients empirically
vary across specific cognitive abilities can be used to
evaluate the explanatory power of theories of intelligence (e.g., Rushton & Jensen, 2010). Toward this end,
we performed a meta-analysis of relevant empirical findings from 23 independent twin studies conducted with
representative samples (combined N = 7,852).
Empirical Data
Method. We first gathered data from previous research
on the relation between subtest g loadings and subtest
heritability coefficients in the Wechsler scales of intelligence (Jensen, 1987, 1998; Rijsdijk, Vernon, & Boomsma,
2002). Next, we conducted a comprehensive search to
locate all studies that involved the Wechsler Intelligence
Scale for Children (WISC), the Wechsler Adult Intelligence
Scale (WAIS), or revisions of either one that provided sufficient information to obtain heritability coefficients on IQsubtest level.
From WISC and WAIS manuals (Wechsler, 1949, 1955,
1974, 1981, 1991, 1992, 1997, 2002, 2005), we obtained
subtest loadings on the first principal component (see
Tables S1 and S2 in the Supplemental Material available
online). The subtest reliability coefficients are provided
in Tables S3 and S4 in the Supplemental Material. The
squared loadings served as approximations of the subtests’ proportions of variance shared with general intelligence. Cultural load was operationalized as the average
proportion of items that were adjusted in each subtest of
the WAIS-III when the scale was adapted for use in 13
countries (Georgas et al., 2003). Because the Wechsler
Verbal IQ (VIQ) and Performance IQ (PIQ) subscales
mapped well on cultural load (see Table 1), we also
searched for all WISC and WAIS studies that reported
heritability coefficients of VIQ and PIQ.
In the analyses, we assumed that tests were independently and randomly sampled from a population of
tests to which the associations generalize. Because this
assumption may be incorrect, inferences may pertain to
the Wechsler scales only. Other limitations concerning
the analyses are addressed in the Conclusion and
Discussion sections.
Analysis 1. Our first analysis involved six WISC ( Jacobs
et al., 2001; LaBuda, DeFries, & Fulker, 1987; Luo, Petrill,
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
Nature and Nurture of Intelligence
2423
Table 1. Cultural Load of the Wechsler Intelligence Test
Subtests on the Verbal IQ (VIQ) and Performance IQ (PIQ)
Subscales
Subtest
Vocabulary
Information
Comprehension
Similarities
Arithmetic
Picture Completion
Picture Arrangement
Block Design
Coding
Digit Span
Object Assembly
Cultural load
(percentage
of adapted items)
Scale
35
22
15
9
8
3
2
1
0
0
0
VIQ
VIQ
VIQ
VIQ
VIQ
PIQ
PIQ
PIQ
PIQ
VIQ
PIQ
5.
Note: Subtest cultural load was obtained from Georgas, van de Vijver,
Weiss, and Saklofske (2003, Table 18.1), except for the Coding subtest
cultural load, which was obtained from F. J. R. van de Vijver (personal
communication, November 30, 2011).
& Thompson, 1994; Owen & Sines, 1970; Segal, 1985;
Williams, 1975) and four WAIS studies (Block, 1965;
Friedman et al., 2008; Rijsdijk, et al., 2002; Tambs, Sundet,
& Magnus, 1984) from which we obtained subtest-level
heritability coefficients (see Tables S5 and S6 in the Supplemental Material). Figure 2 presents these coefficients
and each subtest’s proportion of variance shared with
general intelligence, ranked according to the subtests’
cultural load. Proportion of variance was clearly a function of cultural load: The greater the cultural load, the
greater the squared loading on the first principal component. In the WAIS results, the subtest heritability coefficient also appeared to be a function of cultural load:
The greater the cultural load, the greater the heritability
coefficient. This relation did not appear in the WISC
results.
To test the relationships statistically, and to rule out the
possibility that the relationships were due to differences
in subtest reliability, we examined the WISC and WAIS
data separately, taking the following steps.
1.
2.
3.
4.
We first computed each subtest’s mean loading on
the first principal component and mean reliability
coefficient.
We squared these loadings, which resulted in
proportions of variance shared with general
intelligence.
We divided these proportions by the corresponding mean reliability coefficient, which resulted in
corrected proportions of variance shared with
general intelligence.
6.
7.
8.
9.
We computed the Pearson correlation between
these corrected proportions and the subtests’ cultural loadings and obtained the corresponding
(one-tailed) p value.
We divided, within each of the studies from which
we obtained heritability coefficients, each subtest’s heritability coefficient by the corresponding
mean reliability coefficient, which resulted in heritability coefficients corrected for attenuation.
We computed, within each of these studies, the
Pearson correlation between the subtests’ corrected heritability coefficients (h2) and corrected
proportions of variance shared with general intelligence (gl 2).
We pooled (weighted averaged) these correlations, whereby the square root of the studies’ sample sizes constituted the weights.
We calculated a combined p value using the
Stouffer method (Rosenthal, 1991; i.e., we transformed each one-tailed p value into a z value,
multiplied each z value by the square root of the
corresponding study sample size, summed the
outcomes, divided this sum by the number of
studies, back-transformed the outcome into a onetailed p value, and doubled this one-tailed p value
to obtain a two-tailed p value).
Finally, we repeated Steps 7 and 8 with the correlations between heritability coefficient and cultural load (cl 2).
The correlation between cl 2 and gl 2 was positive, high,
and clearly significant in both the WISC (r = .82, p < .001)
and WAIS (r = .83, p < .001) studies. In the WAIS studies,
the pooled correlations between cl2 and h2 (r = .40, z =
2.65, combined p = .01) and between gl2 and h2 (r = .34,
z = 2.42, combined p = .02) were also positive and significant. In the WISC studies, the pooled correlations
between cl 2 and h 2 (r = .30, z = 1.50, combined p = .15)
and between gl 2 and h 2 (r = .27, z = 1.34, combined p =
.18) were in the same direction, but they did not differ
significantly from 0.
Analysis 2. Our second analysis involved five WISC
and seven WAIS studies that reported heritability coefficients on VIQ and PIQ subscale levels (respectively h2VIQ
and h2PIQ; see Table S7 in the Supplemental Material).
Corroborating the findings presented in the Analysis 1
section, results from a one-tailed, paired Wilcoxon test
showed that in the WAIS studies, h2VIQ was greater than
h2PIQ (median of the differences = 0.07, W = 19, p = .047).
In the WISC studies, there was no such effect (median of
the differences = 0.00, V = 5, p = .5724). Results from the
only longitudinal study in our meta-analysis showed that
the increase in h2VIQ (.84
.46 = .38) was significant,
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
2424
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
Vocabulary
Information
Comprehension
Similarities
Arithmetic
Picture Completion
Picture Arrangement
Block Design
Digit Span
Coding
Object Assembly
Subtest
Subtest
.4
.6
.8
1.0
.2
WISC
.6
Heritability Coefficient
.4
Segal (1985)
Luo et al. (1994)
Williams (1975)
LaBuda et al. (1987)
Jacobs et al. (2001)
.8
1.0
Proportion of Variance Shared With General Intelligence
.2
WISC
Vocabulary
Information
Comprehension
Similarities
Arithmetic
Picture Completion
Picture Arrangement
Block Design
Digit Span
Coding
Object Assembly
Vocabulary
Information
Comprehension
Similarities
Arithmetic
Picture Completion
Picture Arrangement
Block Design
Digit Span
Coding
Object Assembly
.2
.4
.6
WAIS
.8
1.0
.2
WAIS
.6
Heritability Coefficient
.4
Block (1968)
Tambs et al. (1984)
Rijsdijk et al. (2002)
Friedman (2008)
.8
1.0
Proportion of Variance Shared With General Intelligence
WAIS US
WAIS-R US
WAIS-III US
WAIS-III NL
Fig. 2. Subtests’ proportion of variance shared with general intelligence (upper panels) and heritability coefficients (lower panels) of the subtests of the Wechsler Intelligence Scale
for Children (WISC; left panels) and Wechsler Adult Intelligence Scale (WAIS; right panels). Subtests are ranked according to their cultural load (see Table 1). WISC US = Wechsler
(1949); WAIS US = Wechsler (1955); WISC-R US = Wechsler (1974); WAIS-R US = Wechsler (1981); WISC-III US = Wechsler (1991); WISC-III UK = Wechsler (1992); WAIS-III
US = Wechsler (1997); WISC-III NL = Wechsler (2002); WAIS-III NL = Wechsler (2005). Additional data were drawn from the following studies: Block (1965); Friedman et al. (2008);
Jacobs et al. (2001); LaBuda, DeFries, and Fulker (1987); Luo, Petrill, and Thompson (1994); Owen and Sines (1970); Rijsdijk, Vernon, and Boomsma (2002); Segal (1985); Tambs,
Sundet, and Magnus (1984); and Williams (1975).
Vocabulary
Information
Comprehension
Similarities
Arithmetic
Picture Completion
Picture Arrangement
Block Design
Digit Span
Coding
Object Assembly
WISC US
WISC-R US
WISC-III US
WISC-III UK
WISC-III NL
Subtest
Subtest
Nature and Nurture of Intelligence
2425
Minnesota Twin Study
HB Vocabulary
Spelling
.7
Card Rotations
Information
Identical Pictures
Heritability Coefficient
.6
Block Design
Hidden Patterns
Paper Form Board
.5
Raven
Pedigrees
Arithmetic
Subtraction/Multiplication
Coding
Different Uses
Digit Span
Word Fluency
Vocabulary
Word Beginnings/Endings
Spatial Ability
Mental Rotation
Things Categories
Object Assembly
Speed of Closure
Picture Arrangement Comprehension
Mechanical Ability
Inductive Reasoning
Associative Memory
Cubes
Paper Folding
Similarities
Verbal−Proverbs
Memory Span
Meaningful Memory
Verbal−Vocabulary
Numerical Ability
Perceptual Speed
.4
Delayed Visual Memory
Picture Completion
.3
Flexibility of Closure
Lines and Dots
.2
Immediate Visual Memory
.1
.2
.3
Proportion of Variance Shared With General Intelligence
Fig. 3. Heritability coefficients for subtests used in the Minnesota Study of Twins Reared Apart (Johnson et al., 2007) as a function of the proportion
of variance shared with general intelligence.
whereas the increase in h2PIQ (.74 .64 = .10) was not
(see Hoekstra, Bartels, & Boomsma, 2007).
Analysis 3. Our third analysis involved data from the
Minnesota Study of Twins Reared Apart ( Johnson et al.,
2007), in which 126 adult twin pairs completed 42 cognitive subtests from diverse batteries, including the WAIS.
We gathered the subtests’ heritability coefficients (h2) and
computed—on the basis of Johnson et al.’s (2007) factor
analytical results—the subtests’ proportions of variance
explained by the highest-order factor (gl 2). The Pearson
correlation between h2 and gl 2 was positive, of mediumto-large effect size, and significant (r = .50, p < .001).
Figure 3 shows h2 set out against gl 2. The highly culturally loaded Wechsler Arithmetic, Information, and
Vocabulary subtests had relatively large h2 and gl 2 values,
and the same applied to similar tests from other batteries,
such as those involving multiplication and subtraction,
spelling, and vocabulary.
Statistical testing of the relations between cultural load
on the one hand and g loading and heritability on the
other hand required the operationalization of the cultural
load of each subtest. We investigated various operationalizations, but they all were based on Jensen’s (1980)
definition of culture-reduced tests, which are “those that
are nonlanguage and nonscholastic and do not call for
any specific prior information for a plus scored [i.e., correct] response” (p. 374). For example, in one operationalization, subtests were categorized as culture loaded if the
first-order verbal and scholastic factors in combination
explained more variance in the subtest than did all other
first-order factors together. Otherwise, subtests were categorized as culture reduced. In another operationalization, subtests were categorized as culture loaded if they
loaded positively on the second-order verbal factor (on
which the first-order factors verbal, scholastic, fluency,
and number loaded). In a third operationalization, subtests were categorized as culture loaded if, on the basis of
subtest descriptions, it was reasonable to assume that the
completion of the items involved language (as in, e.g.,
the Vocabulary and Verbal-Proverbs subtests), that scholastic skills were measured (as in, e.g., the Arithmetic and
Spelling subtests), or that completion of the items
depended strongly on prior information (as in, e.g., the
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
Kan et al.
2426
Information subtest). Table S8 in the Supplemental
Material shows this information at the subtest level.
Across the different operationalizations of cultural
load (cl 2), the point-biserial correlations between dichotomous cl 2 and continuous gl 2 and h2 were always positive and of medium-to-large effect size. Moreover, they
were almost always significant.
Conclusion
Each subtest’s proportion of variance in IQ shared with
general intelligence was a function of cultural load: The
more culture loaded, the higher this proportion. In addition, in adult samples, culture-loaded tests tended to have
greater heritability coefficients than did culture-reduced
tests, and there was a relationship between subtest’s proportion of variance shared with general intelligence and
heritability. In child samples, these relationships were in
the same direction, but correlations were small and
insignificant.
The interpretation of these results is complicated for at
least two reasons. First, on the one hand, the distribution
of the standard errors of g loadings, heritability estimates,
and cultural loadings are unknown, and data points may
be clustered, which makes the standard errors of the correlations not completely trustworthy. On the other hand,
the power to test the correlations is low, and the risk of
making a Type II error is high, which is one of the reasons alpha levels of .10 are also often considered in these
kinds of analyses ( Jensen, 1998).
Second, a correlation between, for instance, g loading
and heritability coefficient is in line with the hypothesis
that the g factor is the most heritable factor ( Jensen,
1998), but a test of the significance of this correlation
does not provide the means to test whether the g factor
is indeed the most heritable factor1 (Dolan & Hamaker,
2001). The method merely serves to evaluate competing
theories of intelligence (Rushton & Jensen, 2009): A significant correlation denotes that a phenomenon exists
that is in need of theoretical explanation. Theories that
account for the correlation are stronger (with respect to
this correlation) than are theories that do not account for
it or are silent about it. The same line of reasoning holds
for the correlations of cultural load with g loadings and
heritability coefficients. In light of these observations, we
discuss the implications of our results.
Discussion
Our result showing that culture-loaded knowledge
tests (crystallized tests) are more strongly related to general intelligence than are culture-reduced cognitiveprocessing tests (fluid tests) fits better with the idea that
g loadings reflect societal demands (Dickens, 2008) than
that they reflect cognitive demands ( Jensen, 1987).
Furthermore, in adult samples, our finding that the heritability coefficients of culture-loaded tests tend to be
larger than those of culture-reduced tests calls for an
explanation, given that this result does not follow from
the subtest-complexity and investment hypotheses of g
theory and fluid-crystallized theory.
One way to account for the relatively high heritability
coefficients of culture-loaded tests is to assume an equal
genetic effect on cognitive abilities when environmental
variance in highly culturally dependent knowledge
is lower than in less culturally dependent cognitiveprocessing abilities (e.g., because society creates a
homogenous environment for these abilities). This
assumption comes closer to constituting a reformulation
of the effect than to constituting a theoretical explanation, however (e.g., both the nature of g and the way
cognitive abilities are related to this factor therefore still
require explanation). We also doubt that our Western
society creates a homogeneous learning environment,
given that schools and school systems differ strongly.
Another way to account for the relation between heritability coefficient and cultural load, formulated within g
theory (including the investment hypothesis), is to
assume that the acquisition of crystallized abilities is considerably more cognitively demanding than is the solving
of items from even the most complex fluid subtests, such
that crystallized-test performance eventually depends
more strongly on g than does fluid-test performance. This
assumption leaves unanswered the degree of cognitive
demand required in the acquisition of crystalized abilities, such as vocabulary. In addition, the assumption does
not automatically provide an account for the increase in
heritability of IQ (Haworth et al., 2010).
We believe our findings are best understood in
terms of genotype-environment covariance. Because the
acquisition of knowledge depends on cognitive processing, individuals who develop relatively high levels of
cognitive-processing abilities tend to achieve relatively
high levels of knowledge. High achievers are more likely
to end up in cognitively demanding environments that
encourage and facilitate the further development of a
wide range of knowledge and skills. The contents and
organization of these environments largely reflect societal demands. These societal demands thus influence the
degree of dynamical interaction among cognitive processes and knowledge and, hence, their intercorrelations.
In this way, the societal demands determine IQ-subtest
loadings on the general factor of intelligence and, eventually, the degree to which broad-sense heritability coefficients of IQ subtests include the effects of (growing)
genotype-environment covariance. In view of theoretical
parsimony, we conclude that the assumption of a true
causal g can be incorporated but that this is not required.
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
Nature and Nurture of Intelligence
2427
Although we were not able to test the different hypotheses above on the basis of the data presented here, we
maintain that our results are difficult to understand, especially without appreciating the role of culture, education,
and experience in the development of heritable intelligence, if only because of the relationship between heritability and cultural dependence—the more heritable, the
more culture dependent. This relationship sheds new
light on the long-standing nature-versus-nurture debate.
We hope that in future behavior-genetic studies, researchers will model the effects of genotype-environment covariance to test our hypothesis that heritability coefficients
differ across cognitive abilities as a result of differences in
the contribution of genotype-environment covariance,
such that genotype-environment effects are larger on
culture-loaded than on culture-reduced cognitive processing abilities.
Author Contributions
K.-J. Kan, J. M. Wicherts, and H. L. J. van der Maas developed
the study concept, and all authors contributed to the study
design. K.-J. Kan and J. M. Wicherts collected the data. K.-J. Kan
analyzed and interpreted the data under the supervision of
H. L. J. van der Maas and C. V. Dolan. K.-J. Kan and J. M.
Wicherts drafted the manuscript, and C. V. Dolan and H. L. J.
van der Maas provided critical revisions. All authors approved
the final version of the manuscript for submission.
Acknowledgments
We thank Sanne Haring, Denny Borsboom, Paul de Boeck,
Peter Halpin, and Rogier Kievit for their assistance, comments,
and discussion.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
Funding
K.-J. Kan was supported by European Research Council Grant
230374.
Supplemental Material
Additional supporting information may be found at http://pss
.sagepub.com/content/by/supplemental-data
Note
1. To test this, one can employ structural equation modeling.
References
Block, J. B. (1965). Hereditary components in the performance
of twins on the WAIS. Louisville, KY: University of Louisville
School of Medicine.
Cattell, R. B. (1971). Abilities: Their structure, growth, and
action. Boston, MA: Houghton Mifflin.
Dickens, W. T. (2008, June). What is g? Paper presented at
the annual meeting of the Behavior Genetics Association,
Louisville, KY.
Dickens, W. T., & Flynn, J. R. (2001). Heritability estimates versus large environmental effects: The IQ paradox resolved.
Psychological Review, 108, 346–369.
Dolan, C. V., & Hamaker, E. L. (2001). Investigating Black–
White differences in psychometric IQ: Multi-group confirmatory factor analyses of the WISC-R and K-ABC and a
critique of the method of correlated vectors. In F. Columbus
(Ed.), Advances in Psychology Research (Vol. 6, pp. 31–59).
Huntington, NY: Nova Science.
Fancher, R. E. (1996). Pioneers of psychology. New York, NY:
Norton.
Friedman, N. P., Miyake, A., Young, S. E., DeFries, J. C., Corley,
R. P., & Hewitt, J. K. (2008). Individual differences in executive functions are almost entirely genetic in origin. Journal
of Experimental Psychology: General, 137, 201–225.
Georgas, J., van de Vijver, F. J. R., Weiss, L. G., & Saklofske,
D. H. (2003). A cross-cultural analysis of the WISC-III. In
J. Georgas, L. G. Weiss, F. J. R. van de Vijver, & D. H.
Saklofske (Eds.), Culture and children’s intelligence: Crosscultural analysis of the WISC-III (pp. 277–314). Amsterdam,
The Netherlands: Academic Press.
Haworth, C. M. A., Wright, M. J., Luciano, M., Martin, N. G.,
de Geus, E. J., van Beijsterveldt, C. E. M., . . . Plomin, R.
(2010). The heritability of general cognitive ability increases
linearly from childhood to young adulthood. Molecular
Psychiatry, 15, 1112–1120.
Hoekstra, R. A., Bartels, M., & Boomsma, D. I. (2007).
Longitudinal genetic study of verbal and nonverbal IQ
from early childhood to young adulthood. Learning and
Individual Differences, 17, 97–114.
Horn, J. L. (1985). Remodeling old models of intelligence:
Gf-Gc theory. In B. B. Wolman (Ed.), Handbook of intelligence (pp. 267–300). New York, NY: Wiley.
Jacobs, N., Gestel, S., Derom, C., Thiery, E., Vernon, P., Derom,
R., & Vlietinck, R. (2001). Heritability estimates of intelligence in twins: Effect of chorion type. Behavior Genetics,
31, 209–217.
Jensen, A. R. (1980). Bias in mental testing. New York, NY:
Free Press.
Jensen, A. R. (1987). The g beyond factor analysis. In R. R.
Ronning, J. A. Glover, J. C. Conoley, & J. C. Witt (Eds.),
The influence of cognitive psychology on testing (Vol. 3, pp.
87–142). Hillsdale, NJ: Erlbaum.
Jensen, A. R. (1998). The g factor: The science of mental ability.
Westport, CT: Praeger.
Jensen, A. R. (2012). Educational differences. New York, NY:
Taylor & Francis.
Johnson, W., Bouchard, T. J., Jr., McGue, M., Segal, N. L.,
Tellegen, A., Keyes, M., & Gottesman, I. I. (2007). Genetic
and environmental influences on the verbal-perceptualimage rotation (VPR) model of the structure of mental
abilities in the Minnesota Study of Twins Reared Apart.
Intelligence, 35, 542–562.
Johnson, W., Penke, L., & Spinath, F. M. (2011). Heritability
in the era of molecular genetics: Some thoughts for understanding. European Journal of Personality, 25, 254–266.
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014
Kan et al.
2428
LaBuda, M. C., DeFries, J. C., & Fulker, D. W. (1987).
Genetic and environmental covariance structures among
WISC-R subtests: A twin study. Intelligence, 11, 233–
244.
Luo, D., Petrill, S. A., & Thompson, L. A. (1994). An exploration of genetic g: Hierarchical factor analysis of cognitive
data from the Western Reserve Twin Project. Intelligence,
18, 335–347.
Mackintosh, N. J. (1998). IQ and human intelligence. Oxford,
England: Oxford University Press.
Owen, D., & Sines, J. (1970). Heritability of personality in children. Behavior Genetics, 1, 235–248.
Pedersen, N. L., Plomin, R., Nesselroade, J. R., & McClearn,
G. E. (1992). A quantitative genetic analysis of cognitive abilities during the second half of the life span. Psychological
Science, 3, 346–353.
Plomin, R., DeFries, J. C., McClearn, G. E., & McGuffin, P.
(2008). Behavioral genetics. New York, NY: Worth.
Rijsdijk, F. V., Vernon, P. A., & Boomsma, D. I. (2002).
Application of hierarchical genetic models to Raven and
WAIS subtests: A Dutch twin study. Behavior Genetics, 32,
199–210.
Rosenthal, R. (1991). Meta-analytic procedures for social
research. Newbury Park, CA: Sage.
Rushton, J. P., & Jensen, A. R. (2010). The rise and fall of the
Flynn Effect as a reason to expect a narrowing of the
Black–White IQ gap. Intelligence, 38, 213–219.
Scarr, S., & McCartney, K. (1983). How people make their
own environments: A theory of genotype
environment
effects. Child Development, 54, 424–435.
Segal, N. L. (1985). Monozygotic and dizygotic twins: A comparative analysis of mental ability profiles. Child Development,
56, 1051–1058.
Spearman, C. E. (1927). The abilities of man: Their nature and
measurement. New York, NY: Macmillan.
Tambs, K., Sundet, J. M., & Magnus, P. (1984). Heritability analysis of the WAIS subtests: A study of twins. Intelligence, 8,
283–293.
van der Maas, H. L. J., Dolan, C. V., Grasman, R. P. P. P.,
Wicherts, J. M., Huizenga, H. M., & Raijmakers, M. E. J.
(2006). A dynamical model of general intelligence: The positive manifold of intelligence by mutualism. Psychological
Review, 113, 842–861.
Wechsler, D. (1949). Wechsler Intelligence Scale for Children:
Manual. New York, NY: Psychological Corp.
Wechsler, D. (1955). Manual for the Wechsler Adult Intelligence
Scale. New York, NY: Psychological Corp.
Wechsler, D. (1974). Wechsler Intelligence Scale for ChildrenRevised. New York, NY: Psychological Corp.
Wechsler, D. (1981). Wechsler Adult Intelligence Scale-Revised.
San Antonio, TX: Psychological Corp.
Wechsler, D. (1991). Wechsler Intelligence Scale for Children—
Third Edition. San Antonio, TX: Psychological Corp.
Wechsler, D. (1992). Wechsler Intelligence Scale for Children—
Third Edition UK manual. London, England: Psychological
Corp.
Wechsler, D. (1997). Administration and scoring manual:
Wechsler Adult Intelligence Scale—Third Edition. San
Antonio, TX: Psychological Corp.
Wechsler, D. (2002). WISC-III NL handleiding. London, England:
Psychological Corp.
Wechsler, D. (2005). Wechsler Adult Intelligence Scale—III:
Nederlandse bewerking. Technische handleiding. Amsterdam, The Netherlands: Harcourt Assessment B.V.
Williams, T. (1975). Family resemblance in abilities: The
Wechsler scales. Behavior Genetics, 5, 405–409.
Downloaded from pss.sagepub.com by Jelte Wicherts on March 16, 2014